ChatGPT 5.3 release near as model reportedly spotted in A/B tests

Users are being pulled into A/B tests for what looks like GPT-5.3, and this could be a sign the release is very close. X user Kirstin Severino (@KirstinSeverino) spotted the “You’re giving feedback on a new version of ChatGPT” prompt in her app today, suggesting OpenAI is already testing a new model in the wild.

Severino highlighted this test with a comment under a post by AI engineer Dan McAteer, claiming GPT-5.3, codenamed “Garlic,” is dropping this Thursday, February 26.

He says the model scores 83.7% on SimpleBench, a reasoning test designed to catch the kind of obvious common-sense errors most LLMs still make, clearing the human baseline. He’s also claiming it “blows every previous model out of the water on all non-coding benchmarks” and calls it a “GPT-3 to GPT-4 moment” in terms of the leap in quality. But of course, we’d suggest taking that with a grain of salt since we’ll only know for certain once the model actually drops.

This isn’t McAteer’s first intel on Garlic. Back in January, he reposted Sam Altman’s comment confirming the next model will be dubbed GPT 5.3, noting the “Garlic” codename, and describing the model as having stronger pre-training (what insiders call “big model smell”) and IMO Gold-level mathematical reasoning.

The pre-training angle has some backing from inside OpenAI. Mark Chen, the company’s Chief Research Officer, appeared on Ashlee Vance’s Core Memory podcast late last year and hinted at exactly this.

He said OpenAI had been “supercharging pre-training efforts for the last half year” and pushed back on the idea that scaling is dead: “We have algorithmic breakthroughs that enable us to keep scaling the models.” He didn’t name any specific model, but the timing seems to line up.

OpenAI has had a packed few weeks. The company began testing ads inside ChatGPT in the US this month, and before that, it announced plans to retire GPT-4o, which kicked off a whole #Keep4o movement from users who weren’t happy about the change.

No official word from OpenAI on GPT-5.3 yet. But with A/B tests running live and Thursday just around the corner, things are clearly heating up.

We stand out from the tech-media crowd because we break news stories; we mainly bring you stuff that you won’t find anywhere in the mainstream tech media. Our stories have been picked up by some of the world’s most popular websites and media outlets—more info is available here.